We are pleased to announce that the official releases of Update 1 for HPC Pack 2012 R2, and the newest version of Microsoft MPI (MS-MPI) are now available for download. The installation package for HPC Pack 2012 R2 Update 1 is available for download here. As always, HPC Pack includes the latest version of MS-MPI, but in case you need the redistributable installation package for MS-MPI v5, you can download it from here.
The HPC Pack Image in the Azure Gallery, for both Azure Global and Azure China, is also now generally available. Together with this image, an official PowerShell script tool to deploy an HPC Pack cluster in infrastructure as a service (IaaS) VMs is now downloadable from here.
About two months ago, we announced the release of the Preview version of the Update 1 for HPC Pack 2012 R2 and the Beta release of MS-MPI v5. Today, we want to thank everyone who helped us test early releases and sent us feedback. We made some important and necessary changes to our official release as a result.
The following are the significant new features in HPC Pack 2012 R2 Update 1:
A major focus in Update 1 has been to quickly and robustly create a Windows HPC cluster using Azure virtual machines (IaaS). In order to achieve this goal, we released an HPC Pack image in the Azure Virtual Machine Gallery. All customers with an Azure subscription can use the HPC Pack IaaS deployment script tools to quickly deploy an HPC cluster with just a few steps. For details about how to use these tools, you can refer to the help document here.
After a new cluster in Azure has been set up, you can use new tools to manage your HPC cluster in Azure. You can easily create, start, stop, or delete compute nodes as virtual machines on your cluster. Custom compute node images are supported. You can leverage this customization to make your whole deployment process much faster when installing applications and other dependencies. For more details, please see the online help document here.
We also continued the investment around our existing scenario for “bursting” to Azure with Platform as a Service (PaaS) compute instances. To further enhance this scenario, we added grow/shrink capabilities to the HPC Pack scheduler. Now you can configure your HPC cluster to automatically adjust the number of active VMs in Azure, whether using virtual machine compute nodes or PaaS compute nodes. For more details, see the help document here.
There are also several important new support and bug fixes added for HPC Pack, such as support of SQL Server 2014 and moving nodes to a different cluster. For more details, you can refer to the What’s New and Release Notes documents.
MS-MPI v5 includes the following features, improvements, and fixes:
Improved the affinity algorithms for mpiexec.exe (/affinity and /affinity_layout options)
Added support for MPI_Win_allocate_shared and MPI_Win_shared_query
Added support for MPI_Comm_split_type
Added support for MPI_Get_library_version
Added support for configuring the queue depths used for Network Direct connections using the MSMPI_ND_SENDQ_DEPTH and MSMPI_ND_RECVQ_DEPTH environment variables
Removed the dependency on the Visual Studio runtime
Fixed incompatibility with Visual Studio 2012 and 2013 when building mixed Fortran and C++ applications
Added SAL annotations to the APIs to better support static code analysis of MPI applications
Fixed an issue with sentinel values (such as MPI_IN_PLACE) when using upper case Fortran symbols
Refactored debugging support to implement MQS debugging as defined by the MPI Forum
We hope that you enjoy these updates and we look forward to hearing more feedback from you via the Windows HPC Forum, the MPI forum, or you can email use directly to email@example.com and firstname.lastname@example.org. To learn more about MS-MPI, see Microsoft MPI on MSDN.
Finally, Microsoft is attending Supercomputing 2014 (SC14) this week. We prepared several interesting demos in our booth to show some of the new features we are releasing. We invite you to come and meet with us at SC14, and discuss what you want to do with HPC Pack, MS-MPI, Microsoft Azure, and the other technologies that we offer for your compute intensive workloads.
–The HPC Pack and MS-MPI teams at Microsoft